128206 - the euro symbol input from Sun keyboard cannot be display correctly

Status: RESOLVED INVALID

(MailNews Core :: Internationalization, defect)

Description

[Bolian Yin]

In MailNews, open compose to write a email,  input euro symbol from Sun Keyboard
( AltGraph+e), a wrong symbol of "?" is displayed instead of "€".  The same
problem exists in Composer.

Comment 1

[Bolian Yin]

========Information for Reference:  ===========
Latin-1 code of euro:  164 (0xa4)
Unicode of euro:   8364 (0x20ac)

Press AltGraph+e to input euro from Sun Keyboard  The Keycode mozilla get is
164, That is right for euro. I think there are two methods to fix it:

1. in widget/src/gtk/keysym2ucs.c : keysym2ucs function, map 164 to 0x20ac. It
is reasonable, since mozilla change keyboard value to unicdoe here.

2. in intl/uconv/ucvlatin/8859-15.uf : map 0xa4 to 0xa4 instead of fffd. This
can make mozilla find the euro glyph from 8859-1 charset.  

Frank, 8859-15.uf is generated by you from fromu?  Does this file used only in
converting Unicode ==> 8859-15, since there is a 88590-15.ut there.  And do you
 think such change is needed?

I think the first way is better. 

Comment 2

[Bolian Yin]

Attachment: map euro code of 164 to euro unicode of 0x20ac

Comment 3

[Frank Tang]

reassing to ftang

Comment 4

[ji]

Bolian, do you see the same problem with HTML composer (Tasks | Composer)?

Comment 5

[Frank Tang]

>========Information for Reference:  ===========
>Latin-1 code of euro:  164 (0xa4)
>Unicode of euro:   8364 (0x20ac)
>
>Press AltGraph+e to input euro from Sun Keyboard  
why should AltGraph+ e generate Euro sign from sun keyboard? Is there any
documentation said so? what is the following command tell you?

xmodmap -pk | egrep "\(e\)"
in my linux, it show me
26 0x0065 (e) 0x0045 (E) 0x00a4 (currency)
which mean the AltGR+ e should generate currency sign instead of Euro sign.
Does your Sun keyboard report you the same thing. If so, then it mean the
xmodmap are generate XK_currency instead of XK_EuroSign when you press AltGr + e


>The Keycode mozilla get is
>164, That is right for euro. I think there are two methods to fix it:
This is NOT right for euro. 
if you read /usr/X11R6/include/X11/keysymdef.h and search for XK_EuroSign you
will find out it is 
#define XK_EuroSign 0x20ac
in line 2011 in my 1.3 2000/08/18 version . Not sure about which version you got.
This mean the keysym defined for Euro sign is 20ac, not 164. The problem is your
xmodmap generate the wrong value for euro, you should generate 20ac instead of
164. check your xmodmap.

in widget/src/gtk/keysym2ucs.c, we already map 20ac to 20ac
845 { 0x20ac, 0x20ac }, /*                    EuroSign € EURO SIGN */


>1. in widget/src/gtk/keysym2ucs.c : keysym2ucs function, map 164 to 0x20ac. It
>is reasonable, since mozilla change keyboard value to unicdoe here.
This is WRONG. In /usr/X11R6/include/X11/keysymdef.h it 
#define XK_currency 0x0a4
therefore, if you do that, you will break XK_currency which you should NOT.

>2. in intl/uconv/ucvlatin/8859-15.uf : map 0xa4 to 0xa4 instead of fffd. This
>can make mozilla find the euro glyph from 8859-1 charset.  
This is WRONG too. 8859-15.uf map unicode value to ISO-8859-15. However,
ISO-8859-15 do not encode U+00A4 . 

>Frank, 8859-15.uf is generated by you from fromu?  Does this file used only in
>converting Unicode ==> 8859-15, since there is a 88590-15.ut there.  And do you
> think such change is needed?

The real problem is your xmodmap is setting wrong. 
for example, if you have one file tt
keycode 26 = e EuroSign 
and another file tt1
keycode 26 = e 

and if you do
xmodmap tt1
and then type shift + e then you can type euro into mozilla
to change it back
xmodmap tt
and now shift + e generate E instead of euro sign

your proposal are off track. please study the keysym section in X11 programming
first thanks. 

katakai- can you check the xmodmap sun ship with. What is sun's story about euro
sign in solaris for xmodmap ?

I think this a solaris bug, not a mozilla bug. 

Comment 6

[Frank Tang]

>In MailNews, open compose to write a email,  input euro symbol from Sun Keyboard
>( AltGraph+e), a wrong symbol of "?" is displayed instead of "€".

does it display as a question mark ? currency symbol looks like the 00a4
position in http://charts.unicode.org/Web/U0080.html

Comment 7

[Bolian Yin]

> why should AltGraph+ e generate Euro sign from sun keyboard? Is there any
> documentation said so? what is the following command tell you?

http://www.sun.com/software/white-papers/wp-eurocurrency/
This is an white-paper about "Euro Currency Support in the Solaris"
Below is related to the input of Euro in it:

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2.3.1   Keyboard Support

In April 1998, the European Commission (EC) issued recommendations on the
placement of the euro currency symbol on computer keyboards. The EC
recommendation refers to the three main functional levels in keyboard standards:

    * Level 1: Press the "M" key to produce "m" 
    * Level 2: Press SHIFT and "m" to produce "M"
    * Level 3: Press the AltGraph key and "A" on a UK keyboard to produce "à"

The EC proposed both a short-term and long-term solution. The short-term
proposal is to place the euro currency symbol on the "E" key at level 3. This
means that the euro may be generated by pressing two keys: AltGraph and "E". The
EC also recommends that the symbol be engraved on the keytop, which is common
practice for many level 3 characters on European keyboards, such as German.

Sun Keyboard Strategy

Sun has adopted the short-term proposal for placement of the euro currency
symbol on keyboards, meaning the euro will be placed at level 3 and can be
generated by pressing AltGraph+e.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

> xmodmap -pk | egrep "\(e\)"

The output of this in my solaris is:
15  0x0065 (e)    0x0045 (E)        0x00a4 (currency)

If you press AltGraph+e ( within a euro enabled locale, such as fi.ISO8859-15
"Finland", de_DE.ISO8859-15 "Germany") in the dtterm, the € will appear.  It is
the current input method of euro sign in Solaris.  

I think Mozilla should also support it. You can not expect the keysym value to
be 0x20ac now, or the dtterm will have no way to get euro input from keyborad.

Comment 8

[Frank Tang]

>You can not expect the keysym value to be 0x20ac now, 
why not? it is defined in the /usr/X11R6/include/X11/keysymdef.h include file,
right ? IS dtterm X11R6 complient?

>or the dtterm will have no way to get euro input from keyborad.
Why don't you fix dtterm? Are you sure dtterm does not support 20ac as keysym ?
try the following 
1. create one file name tt which have one line:
keycode 15 = e E EuroSign
2. xmodmap tt
and now try it with both mozilla and dtterm.


katakai- why does Sun map keysym 0xa4 to Euro sign?

Comment 9

[Frank Tang]

they are some discussion about this issue on XFree86 1/2002- see

http://xfree86.org/pipermail/i18n/2002-January/002857.html
http://xfree86.org/pipermail/i18n/2002-January/002858.html
http://xfree86.org/pipermail/i18n/2002-January/002859.html

I think this is a invalid bug. Sun's approach to map keysym 0xa4 to euro sign 
are HACKY and will disable the support for keysym 0xa4 for it's design purpose.
This violate the definitation in X11R6 and I think we should not do that. 

Comment 10

[Bolian Yin]

>> Are you sure dtterm does not support 20ac as keysym ?

I tried. In euro sign enabled locale, dtterm does not support 20ac keysym.

Comment 11

[Masaki Katakai]

I'm now looking for this...

I think the reason is that on Solaris 8 we don't have XK_EuroSign def
in keysymdef.h of X11. New Solaris 9 has the definition and I think
we should follow the standard solution...

Anyway, I'll ask to i18n engineer about this.

Comment 12

[Bolian Yin]

Attachment: support AltGraph+e input method of euro

Comment 13

[Bolian Yin]

frank, I add #ifdef SUNOS4 and SOLARIS to the previous patch.  It will be more
reasonable this time. Many users of Solaris will be thankful for that, if we
support AltGraph+e to input euro. It will be better than nothing.

Comment 14

[Bolian Yin]

the second patch should obsolete the first one, who having that permission
please help me to do that.

Comment 15

[Frank Tang]

Bolian: your patch is wrong, reagardless it is #ifdef SOLARIS or not. If
sollaris really want to map that key to Euro sign, fix it in your xlib. don't
propose hacky code in mozilla.

Comment 16

[Bolian Yin]

jdunn, This bug was in the Netscape6.2.2 SUN BRANCH, it should also be in the
NETSCAPE_7_0_OEM_BRANCH, or regression will happen. This patch is important for
SUN Solaris User to input euro in Netscape. It only affect the version on SUN
Solaris Platforms. So, branchOEM+ is needed.  

Thanks,
Bolian

Comment 17

[Hidetoshi Tajima]

About the patch, I don't think mapping AltGraph+e input to EuroSign (0x20ac)
merely with Mozilla is a right thing to do. You are still getting currency
keysym (0xa4) with the rests of desktop applications, which I'm afraid would
make things look very inconsistent to users.

Note that you are still getting 0xa4 from AltGraph+e input even at dtterm of
ISO8859-15 locales. But, it is just
displayed with EuroSign glyph since you're using iso8859-15 encoded font which
maps 0xa4 to EuroSign glyph.

As it stands on curent Solaris8 and 9, a right way to input 0x20a4 with
AltGraph+e is running desktop clients in UTF-8 locales. This works in any
desktop clients consistantly.

Comment 18

[Hidetoshi Tajima]

Bolian: I forgot to ask you one more thing.

The problem you're claiming is specific to ISO8859-15 locales? When you run
Mozilla in ISO8859-15 locale, you 
get international currency sign displayed for 0xa4 while 
you get EuroSign displayed on other apps. It that the claim?

If so, it sounds like it's possibily a charset ordering issue of fonts rather
than input. But, I don't think it's the design of Mozilla text shaper to change
ISO8859-X
fonts for ASCII/Latin1-supplement code range according to 
the locale's encoding. Frank?

Comment 19

[Bolian Yin]

Hidetoshi, Thanks for evaluation.
It seems that Frank insist that 0x20ac should be generated in order to input
euro from Keyboard (see comment #5). Frank?  If we generate 0xa4 (currency) from
AltGraph+e, and map it into euro when needed according to LOCALE, like other
applications in Solaris do, is that acceptable for Mozilla?  Does Mozilla deal
with the LOCALE issue, how? Thanks.

Comment 20

[Hidetoshi Tajima]

>It seems that Frank insist that 0x20ac should be generated 
> in order to input euro from Keyboard (see comment #5). 

Please note that I agree with Frank on this point above
regarding to "input". You should not turn 0xa4 keysym to anything else in
Mozilla. This kind of thing is not done
in any other application on Solaris.

> Frank?  If we generate 0xa4 (currency) from
> AltGraph+e, and map it into euro when needed according
> to LOCALE, like other applications in Solaris do, is 
> that acceptable for Mozilla?  Does Mozilla deal
> with the LOCALE issue, how? Thanks.

Note that "map 0xa4 into euro on Solaris apps" is done on
font level, but not on input level, and it is a specific
thing when you run apps on ISO8859-15 locales.

And, what I mentioned "possibilly a charset ordering 
issue.." in my last comment is about font mapping but
not about input. I said we *might* use ISO8859-15 fonts
(where 0xa4 has Euro glyph) to draw(and print)
ASCII/Latin-1 (0x20-0x7f, 0x81-0xff) *only when* Mozilla
is running in ISO8859-15 locales.

(I'm not insisting to do this, but just introduce it
a possible option as a live example, which is actually taken on Solaris
ISO8859-15 locales.)

IMO: Mozilla's Unicode based, so character to be displayed 
should be subject to how Unicode book defines, where 0xa4
is CURRENCY SIGN and 0x20ac is EURO SIGN - and that's it. Introducing locale
things to character renderng would make 
things more complicated than necessary.

Comment 21

[Bolian Yin]

OK, then the most direct and effective way is to file a bug in Solaris, Solaris
should generate 0x20ac not 0xa4 keysym whenener a Euro is wanted, whether it use
AltGraph+e or others.  
Then this bug can be closed here. Hidetoshi,  Is that OK?

Comment 22

[Bolian Yin]

Solaris Xlib should fix this. bugtraq 4730547